Research And Implementation Of Intra-domain Traffic Engineering For Failure Recovery

With the rapid development of the Internet technology, there are massive network applications emerging in the Internet, and the Internet needs undertake more and more network traffic and service tasks, which forwards a giant challenge to the network reliability. Traditional IP-based distributed networks might fail frequently because of their own shortcomings, which seriously affected the normal operations of networks as well as the benefits of the network users and network service providers.Related researchers have made a series of relevant researches on the network failure recovery, but past research mainly based on the network connectivity, transferred the network flow from fault paths to recover paths, didn't take into account the network load balance, which easily causes new network congestion. The research about the traffic engineering can improve utilization of network resources, reduce the network link maximum utilization, and make sure the load balance of network traffic, but cannot guarantee the real-time response of network failure. With the development of the network system structure, the concept of SDN forwarded a new idea for us. In this paper, based on the SDN centralized routing architecture, we proposed a failure recovery-oriented intra-domain traffic engineering system structure, and a corresponding online optimization algorithm MCPF. Our work can plan the flow request in the network according to the actual network conditions, make sure the network load balance, as well as realize quickly path optimization, fault recovery and make sure the traffic load balance after fault when a failure occurs in the network. The major work of this paper is as follows:First, we proposed a failure recovery-oriented intra-domain traffic engineering system structure based on the SDN centralized routing architecture, which effectively combined the fault recovery technology and traffic engineering technology together.Second, we presented an online load balancing algorithm MCPF. The algorithm was based on the idea of Dijkstra single-source shortest path, had the advantage of low time complexity. Besides, the algorithm can realize quick path plan of flow requests according to the actual operation situation of network, which can effectively improve the network resources utilization, reduce the maximum link utilization of network links, realize the load balance of network flow and reduce the network congestion.Third, we adopted the BFD fault detection method, which can quickly detect the network link failures. Besides, we improved the BFD detection. When a network forwarding node did not receive a message for successive two times, a failure may be occurred in the network link. In this case the forwarding node noticed the control unit, and the control unit computed all the flow paths in the network link in advance. When the BFD detected a failure, the control unit sent and deployed the recovery paths that the failure link may affected to the network, which can realize quick fault detect and failure recovery, effectively reduce the loss caused by the failure.Fourth, the forwarding units adopted the method of source routing when they transmitted data which along the flow path that control unit specified. Hence we can effectively control the data flow, and the control unit need only distribute the configuration information to source nodes of the flow when we need distribute and deploy the flow paths.